U.S. patent application number 17/019898 was filed with the patent office on 2021-01-07 for non-regular electrical stimulation patterns designed with a cost function for treating neurological disorders.
The applicant listed for this patent is Duke University. Invention is credited to David T. Brocker, Warren M. Grill.
Application Number | 20210001127 17/019898 |
Document ID | / |
Family ID | |
Filed Date | 2021-01-07 |
View All Diagrams
United States Patent
Application |
20210001127 |
Kind Code |
A1 |
Grill; Warren M. ; et
al. |
January 7, 2021 |
Non-Regular Electrical Stimulation Patterns Designed with a Cost
Function for Treating Neurological Disorders
Abstract
Systems and methods for stimulation of neurological tissue
generate stimulation trains with temporal patterns of stimulation,
in which the interval between electrical pulses (the inter-pulse
intervals) changes or varies over time. Compared to conventional
continuous, high rate pulse trains having regular (i.e., constant)
inter-pulse intervals, the non-regular (i.e., not constant) pulse
patterns or trains that embody features of the invention provide a
lower average frequency.
Inventors: |
Grill; Warren M.; (Chapel
Hill, NC) ; Brocker; David T.; (Cary, NC) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Duke University |
Durham |
NC |
US |
|
|
Appl. No.: |
17/019898 |
Filed: |
September 14, 2020 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
15436948 |
Feb 20, 2017 |
10773084 |
|
|
17019898 |
|
|
|
|
15005636 |
Jan 25, 2016 |
9572988 |
|
|
15436948 |
|
|
|
|
14583932 |
Dec 29, 2014 |
9242095 |
|
|
15005636 |
|
|
|
|
13770731 |
Feb 19, 2013 |
8923981 |
|
|
14583932 |
|
|
|
|
12587295 |
Oct 5, 2009 |
8447405 |
|
|
13770731 |
|
|
|
|
16145785 |
Sep 28, 2018 |
|
|
|
12587295 |
|
|
|
|
14542785 |
Nov 17, 2014 |
10086204 |
|
|
16145785 |
|
|
|
|
13897504 |
May 20, 2013 |
9259579 |
|
|
14542785 |
|
|
|
|
12587295 |
Oct 5, 2009 |
8447405 |
|
|
13897504 |
|
|
|
|
61600264 |
Feb 17, 2012 |
|
|
|
61102575 |
Oct 3, 2008 |
|
|
|
61102575 |
Oct 3, 2008 |
|
|
|
Current U.S.
Class: |
1/1 |
International
Class: |
A61N 1/36 20060101
A61N001/36; A61N 1/05 20060101 A61N001/05 |
Claims
1. A neural stimulation device comprising: a pulse generator; at
least one electrode operatively coupled with the pulse generator,
wherein the pulse generator transmits electrical stimulation
comprising temporal patterns to neurological tissue through the at
least one electrode, said temporal patterns including: a first
temporal pattern having a first non-regular pulse train comprising
a plurality of pulses having differing inter-pulse intervals
between pulses to treat a neurological condition, said first
temporal pattern adapted from applying; and subsequent temporal
patterns having a subsequent non-regular pulse train comprising a
plurality of pulses having non-regular inter-pulse intervals
between the pulses to treat a neurological condition, wherein the
subsequent temporal patterns are: (a) adapted from applying a
model-based optimization technique, and (b) differing in their
nature.
2. The neural stimulation device of claim 1, wherein the pulse
generator is configured to apply at least one of the first and
subsequent temporal patterns of stimulation in repeating
succession, whereby the first temporal pattern is different from
the subsequent temporal patterns.
3. The neural stimulation device of claim 2 further comprising an
output port in the pulse generator configured to operatively attach
the at least one electrode.
4. The neural stimulation device of claim 3, wherein the at least
one electrode is an implanted lead.
5. The neural stimulation device of claim 1, wherein the
model-based optimization technique includes applying at least one
of: a genetic algorithm, swarm intelligence algorithms, and
metaheuristic.
6. A method for stimulation of a targeted neurological tissue
region comprising the steps of: applying electrical current to a
targeted neurological tissue region in a brain of an animal using a
pulse generator operatively coupled with at least one electrode,
according to a first non-regular pulse train comprising a plurality
of pulses having differing inter-pulse intervals between the pulses
to treat a neurological condition; applying a model-based
optimization technique creating a subsequent non-regular pulse
train comprising a plurality of second pulses having non-regular
inter-pulse intervals between the second pulses; and applying
electrical current to the targeted neurological tissue region of
the animal using the pulse generator and through the at least one
electrode according to the subsequent non-regular pulse train.
7. The method of claim 6 further comprising repeating the applying
electrical current to the targeted neurological tissue region of
the animal using the pulse generator according to the subsequent
non-regular pulse train in succession, wherein the subsequent
non-regular pulse train is different from the first non-regular
pulse train.
8. The method of claim 6, wherein the model-based optimization
technique includes applying at least one of: a genetic algorithm,
swarm intelligence algorithms and metaheuristic.
9. The method of claim 6 further comprising analyzing results of
the first non-regular pulse train by quantitatively assessing the
first non-regular pulse train having an efficiency measure, E, and
an efficacy measure, S.
10. The method of claim 9, wherein analyzing the results of the
first non-regular pulse train includes applying a cost function (C)
for the first non-regular pulse train based upon E and S, the cost
function weighting E and S differentially to minimize E and S.
11. The method of claim 9, wherein analyzing the results of the
first non-regular pulse train includes applying the cost function
to evaluate the cost of the first non-regular pulse train.
12. The method of claim 9, wherein E is an average stimulation
pulse rate.
13. The method of claim 9, wherein S is a rate or a pattern of
neural activity.
14. The method of claim 9, wherein S is a rate or a pattern of
neural activity.
15. A method of treating Parkinson's disease in a human being
comprising: applying non-regular patterns of electrical stimulation
having a frequency range between 100 Hz and 200 Hz to a human being
via an implanted lead in the human being's thalamus, subthalamus,
or globus pallidus; modeling oscillatory or synchronous activity in
the human being; and generating subsequent non-regular patterns of
electrical stimulation based on the modeled activity and applying
the subsequent non-regular patterns of electrical stimulation to
the human being via the implanted lead, wherein the subsequent
non-regular patterns of electrical stimulation comprise a plurality
of pulses having non-regular, non-random, differing inter-pulse
intervals between the pulses.
16. The method of claim 15 wherein the frequency range is 130
Hz.
17. The method of claim 15 further comprising quantitatively
assessing how thalamus transmitted motor commands are generated in
the human being based upon the applied non-regular patterns of
electrical stimulation prior to generating subsequent non-regular
patterns of electrical stimulation.
18. The method of claim 17 wherein the quantitatively assessing
includes an error function reflecting how faithfully the thalamus
transmitted motor commands are generated by inputs from the cortex
of the human being.
19. The method of claim 17 wherein the quantitatively assessing
includes optimizing stimulation patterns utilizing a cost
function.
20. The method of claim 15 wherein the modeling employs a genetic
algorithm.
21. The method of claim 20 wherein the genetic algorithm is coupled
to a computational model of stimulation of the subthalamic
nucleus.
22. The method of claim 15 wherein the modeling reflects neuronal
firings of neurons in each of the subthalamic nucleus, the global
pallidus exterior, and the global pallidus interior.
23. A method of treating Parkinson's disease in a human being
comprising: applying non-regular patterns of electrical stimulation
having a frequency range of between 100 Hz and 200 Hz to a human
being having Parkinson's disease via an implanted lead in the human
being's thalamus, subthalamus, or globus pallidus; measuring the
oscillatory or synchronous activity in the human being;
quantitatively assessing how thalamus transmitted motor commands
are generated based upon the applied non-regular patterns of
electrical stimulation; and generating subsequent non-regular
patterns of electrical stimulation based on the measured activity
wherein the subsequent non-regular patterns of electrical
stimulation comprise a plurality of pulses having non-regular,
non-random, differing inter-pulse intervals between the pulses.
24. The method of claim 23, further comprising applying the
subsequent non-regular patterns of electrical stimulation to the
human being via the implanted lead reducing symptoms of Parkinson's
disease.
25. The method of claim 24 further comprising quantitatively
assessing how thalamus transmitted motor commands are generated in
the human being based upon the applied non-regular patterns of
electrical stimulation prior to generating the subsequent
non-regular patterns of electrical stimulation.
26. The method of claim 25 wherein the quantitatively assessing
step includes an error function reflecting how faithfully the
thalamus transmitted motor commands are generated by inputs from
the cortex of the human being.
27. The method of claim 26 wherein the quantitatively assessing
comprising optimizing stimulation patterns that comprise a cost
function.
28. The method of claim 25 further comprising modeling the
oscillatory or synchronous activity in the human being via a
genetic algorithm and wherein the subsequent non-regular patterns
of electrical stimulation are based upon both the measured activity
and the modeled activity.
29. The method of claim 28, wherein the genetic algorithm is
coupled to a computational model of stimulation of the subthalamic
nucleus.
30. The method of claim 28 further comprising modeling reflective
of neuronal firings of neurons in each of the subthalamic nucleus,
the global pallidus exterior, and the global pallidus interior.
31. A method of treating epilepsy or movement disorders in a human
being comprising: applying non-regular patterns of electrical
stimulation having a frequency range between 100 Hz and 200 Hz to a
human being having epilepsy or movement disorder via an implanted
lead in the human being's thalamus, subthalamus, or globus
pallidus; modeling and/or measuring the oscillatory or synchronous
activity in the human being; quantitatively assessing how thalamus
transmitted motor commands are generated based upon the applied
non-regular patterns of electrical stimulation; and generating
subsequent non-regular patterns of electrical stimulation based on
the modeling and/or measured activity wherein the subsequent
non-regular patterns of electrical stimulation comprise a plurality
of pulses having non-regular, non-random, differing inter-pulse
intervals between the pulses.
32. The method of claim 31 further comprising applying the
subsequent non-regular patterns of electrical stimulation to the
human being via the implanted lead,
33. The method of claim 32 wherein the frequency range is 130
Hz.
34. The method of claim 32 wherein modeling is performed and
includes a genetic algorithm.
35. The method of claim 34 wherein the genetic algorithm is coupled
to a computational model of stimulation of the subthalamic nucleus.
Description
CROSS-REFERENCE TO RELATED APPLICATION
[0001] This application is a continuation of U.S. patent
application Ser. No. 15/436,948 filed on Feb. 20, 2017, and
entitled "A Neural Stimulation Device with Non-Regular Stimulation
Patterns Designed with a Cost Function for Treating Neurological
Disorders," which is a Continuation of U.S. patent application Ser.
No. 15/005,636, filed on Jan. 25, 2016, granted as U.S. Pat. No.
9,572,988, and entitled "A Neural Stimulation Device with
Non-Regular Stimulation Patterns Designed with a Cost Function for
Treating Neurological Disorders," which is a continuation of U.S.
patent application Ser. No. 14/583,932, filed on Dec. 29, 2014,
granted as U.S. Pat. No. 9,242,095, and entitled "A Neural
Stimulation Device with Non-Regular Stimulation Patterns Designed
with a Cost Function for Treating Neurological Disorders," which is
a continuation of U.S. patent application Ser. No. 13/770,731,
filed on Feb. 19, 2013, granted as U.S. Pat. No. 8,923,981, and
entitled "Non-Regular Electrical Stimulation Patterns Designed With
a Cost Function for Treating Neurological Disorders," which claims
the benefit of U.S. Provisional Application Ser. No. 61/600,264,
entitled "Non-Regular Electrical Stimulation Patterns for Treating
Neurological Disorders" filed on Feb. 17, 2012 and U.S. patent
application Ser. No. 13/770,731 is a continuation-in-part of
co-pending U.S. patent application Ser. No. 12/587,295, granted as
U.S. Pat. No. 8,447,405, filed Oct. 5, 2009, and entitled
"Non-Regular Electrical Stimulation Patterns for Treating
Neurological Disorders," which claimed the benefit of U.S.
Provisional Patent Application Ser. No. 61/102,575, filed Oct. 3,
2008, and entitled "Stimulation Patterns For Treating Neurological
Disorders Via Deep Brain Stimulation," which are all incorporated
herein by reference. This application is also a continuation in
part of U.S. patent application Ser. No. 16/145,785, filed on Sep.
28, 2018 and entitled "NON-REGULAR ELECTRICAL STIMULATION PATTERNS
FOR TREATING NEUROLOGICAL DISORDERS" which is a continuation of
U.S. patent application Ser. No. 14/542,785, now U.S. Pat. No.
10,086,204, filed on Nov. 17, 2014, and entitled "NON-REGULAR
ELECTRICAL STIMULATION PATTERNS FOR TREATING NEUROLOGICAL
DISORDERS" which is a continuation of U.S. patent application Ser.
No. 13/897,504, now U.S. Pat. No. 9,259,579, filed May 20, 2013,
and entitled "Non-Regular Electrical Stimulation Patterns for
Treating Neurological Disorders," which is a continuation of U.S.
patent application Ser. No. 12/587,295, now U.S. Pat. No.
8,447,405, filed Oct. 5, 2009, and entitled "Non-Regular Electrical
Stimulation Patterns for Treating Neurological Disorders," which
claimed the benefit of U.S. Provisional Patent Application Ser. No.
61/102,575, filed Oct. 3, 2008, and entitled "Stimulation Patterns
For Treating Neurological Disorders Via Deep Brain Stimulation,"
which are all incorporated herein by reference.
FIELD OF INVENTION
[0002] This invention relates to systems and methods for
stimulating nerves in animals, including humans.
BACKGROUND
[0003] Deep Brain Stimulation (DBS) has been found to be successful
in treating a variety of brain-controlled disorders, including
movement disorders. Generally, such treatment involves placement of
a DBS type lead into a targeted region of the brain through a burr
hole drilled in the patient's skull, and the application of
appropriate stimulation through the lead to the targeted
region.
[0004] Presently, in DBS, beneficial (symptom-relieving) effects
are observed primarily at high stimulation frequencies above 100 Hz
that are delivered in stimulation patterns or trains in which the
interval between electrical pulses (the inter-pulse intervals) is
constant over time. The trace of a conventional stimulation train
for DBS is shown in FIG. 2. The beneficial effects of DBS on motor
symptoms are only observed at high frequencies, while low frequency
stimulation has generally been thought to exacerbate symptoms. For
instance, thalamic DBS at less than or equal to 50 Hz has been
shown to increase tremor in patients with essential tremor.
Similarly, 50 Hz DBS has been shown to produce or induce tremor in
pain patients receiving simulation of the ventral posterior medial
nucleus of the thalamus (VPM), but the tremor disappears when the
frequency is increased. Likewise, DBS of the subthalamic nucleus
(STN) at 10 Hz has been shown to worsen akinesia in patients with
Parkinson's disease (PD), while DBS at 130 Hz leads to significant
improvement in motor function. Similarly, relatively high frequency
stimulation of the globus pallidus (GP) at or above 130 Hz has been
shown to improve dystonia, whereas stimulation at either 5 or 50 Hz
may lead to significant worsening.
[0005] Model studies also indicate that the masking of pathological
burst activity occurs only with sufficiently high stimulation
frequencies. See Grill et al. 2004, FIG. 1. Responsiveness of
tremor to changes in DBS amplitude and frequency are strongly
correlated with the ability of applied stimuli to mask neuronal
bursting. See Kuncel et al. 2007, FIG. 2.
[0006] Although effective, conventional high frequency stimulation
generates stronger side-effects than low frequency stimulation, and
the therapeutic window between the voltage that generates the
desired clinical effect(s) and the voltage that generates undesired
side effects decreases with increasing frequency. Precise lead
placement therefore becomes important. Further, high stimulation
frequencies increase power consumption. The need for higher
frequencies and increased power consumption shortens the useful
lifetime and/or increases the physical size of battery-powered
implantable pulse generators. The need for higher frequencies and
increased power consumption requires a larger battery size, and/or
frequent charging of the battery, if the battery is rechargeable,
or replacement of the battery if it is not rechargeable.
SUMMARY
[0007] The invention provides stimulation patterns or trains with
different temporal patterns of stimulation than conventional
stimulation trains. The invention also provides methodologies to
identify and characterize stimulation patterns or trains that
produce desired relief of symptoms, while reducing the average
stimulation frequency.
[0008] According to one aspect of the invention, the intervals
between stimulation pulses in a pulse pattern or train (in
shorthand called "the inter-pulse intervals") is not constant over
time, but changes or varies over time. These patterns or trains are
consequently called in shorthand "non-regular." According to this
aspect of the invention, the non-regular (i.e., not constant) pulse
patterns or trains provide a lower average frequency for a given
pulse pattern or train, compared to conventional continuous, high
rate pulse trains having regular (i.e., constant) inter-pulse
intervals. Having a lower average frequency, the non-regular
stimulus patterns or trains make possible an increase in the
efficacy of stimulation by reducing the intensity of side effects;
by increasing the dynamic range between the onset of the desired
clinical effect(s) and side effects (and thereby reducing
sensitivity to the position of the lead electrode); and by
decreasing power consumption, thereby providing a longer useful
battery life and/or a smaller implantable pulse generator, allowing
battery size reduction and/or, for rechargeable batteries, longer
intervals between recharging.
[0009] The non-regular stimulation patterns or trains can be
readily applied to deep brain stimulation, to treat a variety of
neurological disorders, such as Parkinson's disease, movement
disorders, epilepsy, and psychiatric disorders such as
obsessive-compulsion disorder and depression. The non-regular
stimulation patterns or trains can also be readily applied to other
classes electrical stimulation of the nervous system including, but
not limited to, cortical stimulation, spinal cord stimulation, and
peripheral nerve stimulation (including sensory and motor), to
provide the attendant benefits described above and to treat
diseases such as but not limited to Parkinson's Disease, Essential
Tremor, Movement Disorders, Dystonia, Epilepsy, Pain, psychiatric
disorders such as Obsessive Compulsive Disorder, Depression, and
Tourette's Syndrome.
[0010] According to another aspect of the invention, systems and
methodologies make it possible to determine the effects of the
temporal pattern of DBS on simulated and measured neuronal
activity, as well as motor symptoms in both animals and humans. The
methodologies make possible the qualitative determination of the
temporal features of stimulation trains.
[0011] The systems and methodologies described herein employ a
genetic algorithm, coupled to a computational model of DBS of the
STN, to develop non-regular patterns of stimulation that produced
efficacy (as measured by a low error function, E) at lower
stimulation frequencies, F. The error function, E, is a
quantitative measure from the model which assesses how faithfully
the thalamus transmitted motor commands that are generated by
inputs from the cortex. A very high correlation exists between E
and symptoms in persons with PD, and therefore E is a valid
predictor for the efficacy of a stimulation train in relieving
symptoms (see Dorval et al., 2007).
[0012] Previous efforts (see Feng et al. 2007) sought to design
stimulation trains that minimized the total current injection. The
systems and methodologies disclosed herein include an objective
function that maximizes therapeutic benefit (by minimizing the
error function) and improves stimulation efficiency (by reducing
the stimulation frequency), using a model of the STN that
reproduces the frequency tuning of symptom reduction that has been
documented clinically. In contrast, the Feng et al. model showed,
incorrectly, symptom reduction with regular, low frequency
stimulation. The inventors have identified novel non-regular
temporal patterns of stimulation, while Feng et al. identified
regular low frequency (.about.10 Hz) trains that previous clinical
work has demonstrated to be ineffective.
[0013] A neural stimulation device may include a pulse generator
configured to transmit a first temporal pattern of stimulation for
application to neurological tissue having a first non-regular pulse
train, the first non-regular pulse train including a first
plurality of single pulses (first singlets) and embedded first
multiple pulse groups (first n-lets), with non-regular inter-pulse
intervals between the first singlets and first n-lets, as well as
non-regular inter-pulse intervals within the first n-lets
themselves. The pulse generator may also be configured to transmit
a second temporal pattern of stimulation for application to
neurological tissue having a second non-regular pulse train, the
second non-regular pulse train including a second plurality of
single pulses (second singlets) and embedded second multiple pulse
groups (second n-lets), with non-regular inter-pulse intervals
between second singlets and second n-lets, as well as non-regular
inter-pulse intervals within the second n-lets themselves, the
second temporal pattern adapted from applying a model-based
optimization technique after application of the first temporal
pattern of stimulation.
[0014] A method for stimulation of a targeted neurological tissue
region may include the steps of applying electrical current to a
targeted neurological tissue region of an animal using a pulse
generator according to a first non-regular pulse train including a
first plurality of single pulses (first singlets) and embedded
first multiple pulse groups (first n-lets), with non-regular
inter-pulse intervals between the first singlets and first n-lets,
as well as non-regular inter-pulse intervals within the first
n-lets themselves, and analyzing results of the first non-regular
pulse train. The method may further include the steps of applying a
model-based optimization technique determining a second non-regular
pulse trains including a second plurality of single pulses (second
singlets) and embedded second multiple pulse groups (second
n-lets), with non-regular inter-pulse intervals between second
singlets and second n-lets, as well as non-regular inter-pulse
intervals within the second n-lets themselves, and applying
electrical current to the targeted neurological tissue region of
the animal using the pulse generator according to the second
non-regular pulse train.
[0015] A neural stimulation device may include a pulse generator
configured to apply a first non-regular pulse train, having at
least one first singlet spaced apart by a first inter-pulse singlet
interval and at least one first n-let having, for each n-let, two
or more pulses spaced apart by a first inter-pulse interval that is
less than the first singlet inter-pulse interval. The pulse
generator may also be configurable to apply a second non-regular
pulse train, having at least one second singlet spaced apart by a
second inter-pulse singlet interval and at least one second n-let
having, for each n-let, two or more pulses spaced apart by a second
inter-pulse interval that is less than the second singlet
inter-pulse interval, the second non-regular pulse trail based upon
an analysis of the first non-regular pulse train. The neural
stimulation device may also include at least one input configured
to operatively connect with at least one electrode.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Operation of the invention may be better understood by
reference to the detailed description taken in connection with the
following illustrations, wherein:
[0017] FIG. 1 is an anatomic view of a system for stimulating
tissue of the central nervous system that includes an lead
implanted in brain tissue coupled to a pulse generator that is
programmed to provide non-regular (i.e., not constant) pulse
patterns or trains, in which the interval between electrical pulses
(the inter-pulse intervals) changes or varies over time.
[0018] FIG. 2 is a diagrammatic trace that shows a conventional
regular high frequency stimulation train, in which the interval
between electrical pulses (the inter-pulse intervals) is
constant.
[0019] FIG. 3 is a diagrammatic trace showing a representative
example of a repeating non-regular pulse pattern or train in which
the inter-pulse intervals are linearly cyclically ramped over
time.
[0020] FIGS. 4 and 5 are diagrammatic traces showing other
representative examples of repeating non-regular pulse patterns or
trains comprising within, a single pulse train, a combination of
single pulses (singlets) and embedded multiple pulse groups
(n-lets), with non-regular inter-pulse intervals between singlets
and n-lets as well as non-regular inter-pulse intervals within the
multiple pulse n-lets.
[0021] FIG. 6 is a reproduction of Example FIG. 1, as described
below.
[0022] FIG. 7 is a reproduction of Example FIG. 2, as described
below.
[0023] FIG. 8 is a reproduction of Example FIG. 3, as described
below.
[0024] FIG. 9 is a reproduction of Example FIG. 4, as described
below.
[0025] FIG. 10 is a reproduction of Example FIG. 5, as described
below.
[0026] FIG. 11 is a raster of the neuronal firings of ten neurons
in each of the subthalamic nucleus, the global pallidus exterior,
and the global pallidus interior, generated by a computer model of
a healthy human.
[0027] FIG. 12 is a raster of the neuronal firings of ten neurons
in each of the subthalamic nucleus, the global pallidus exterior,
and the global pallidus interior, generated by a computer model of
a human having a neurological condition, such as Parkinson's
Disease.
[0028] FIG. 13 is a raster of the neuronal firings of ten neurons
in each of the subthalamic nucleus, the global pallidus exterior,
and the global pallidus interior, generated by a computer model of
an electrical deep brain stimulation of regular interval applied to
the subthalamic nucleus at 45 Hertz.
[0029] FIG. 14 is a raster of the neuronal firings of ten neurons
in each of the subthalamic nucleus, the global pallidus exterior,
and the global pallidus interior, generated by a computer model of
an electrical deep brain stimulation of regular interval applied to
the subthalamic nucleus at 100 Hertz.
[0030] FIG. 15 is a plot of a computer-generated error fraction of
a regular interval electrical DBS applied at the given average
frequency.
[0031] FIG. 16 diagrammatically illustrates an embodiment of a
computer model that may be used to analyze and generate embodiments
stimulation patterns according to the present invention.
[0032] FIG. 17 is a diagram of a general genetic algorithm
process.
[0033] FIG. 18 is an embodiment of a generational crossover of
stimulation patterns according to the present invention.
[0034] FIG. 19 is a plot of a percent decrease cost function versus
the number of generations run in an evolutionary algorithm
according to the present invention.
[0035] FIG. 20 is a raster of the neuronal firings of ten neurons
in each of the subthalamic nucleus, the global pallidus exterior,
and the global pallidus interior, generated by a computer model of
an electrical deep brain stimulation applied to the subthalamic
nucleus according to a first embodiment of a stimulation pattern
according to the present invention.
[0036] FIG. 21 is a raster of the neuronal firings of ten neurons
in each of the subthalamic nucleus, the global pallidus exterior,
and the global pallidus interior, generated by a computer model of
an electrical deep brain stimulation applied to the subthalamic
nucleus according to a second embodiment of a stimulation pattern
according to the present invention.
[0037] FIG. 22 is a raster of the neuronal firings of ten neurons
in each of the subthalamic nucleus, the global pallidus exterior,
and the global pallidus interior, generated by a computer model of
an electrical deep brain stimulation applied to the subthalamic
nucleus according to a third embodiment of a stimulation pattern
according to the present invention.
[0038] FIG. 23 is FIG. 15, further including a plot of the computer
modeled error fractions generated by the use of the stimulation
patterns of FIGS. 20-22.
[0039] FIG. 24 is a graph of a quantitative measurement of the
performance of the stimulation pattern of FIG. 20 as compared to
other stimulation patterns in two human patients which had been
diagnosed with Parkinson's Disease.
[0040] FIG. 25 is a graph of a quantitative measurement of the
performance of the stimulation pattern of FIG. 20 as compared to
other stimulation patterns in a human patient which had been
diagnosed with Parkinson's Disease and had tremor as a primary
motor impairment related thereto.
[0041] FIG. 26 includes graphs in the frequency domain of an
average power of ten GPi neuronal firing sequences across a single
iteration of the indicated stimulation, where such sequences were
computer model generated.
[0042] FIG. 27 is a tuning curve indicating the average power of
ten GPi neuronal firing sequences across ten iterations of the
indicated stimulation, where such sequences were computer model
generated.
[0043] FIG. 28A is a graphical representation of a measure of an
experimental tremor during a trial.
[0044] FIG. 28B is a graphical representation of amplitude of
tremors with a power spectral density calculated for each of the
measured amplitudes.
DETAILED DESCRIPTION
[0045] Reference will now be made in detail to exemplary
embodiments of the present invention, examples of which are
illustrated in the accompanying drawings. It is to be understood
that other embodiments may be utilized and structural and
functional changes may be made without departing from the
respective scope of the invention. Moreover, features of the
various embodiments may be combined or altered without departing
from the scope of the invention. As such, the following description
is presented by way of illustration only and should not limit in
any way the various alternatives and modifications that may be made
to the illustrated embodiments and still be within the spirit and
scope of the invention.
[0046] FIG. 1 is a system 10 for stimulating tissue of the central
nervous system. The system may include a lead 12 placed in a
desired position in contact with central nervous system tissue. In
the illustrated embodiment, the lead 12 may be implanted in a
region of the brain, such as the thalamus, subthalamus, or globus
pallidus for the purpose of deep brain stimulation. However, it
should be understood, the lead 12 may be implanted in, on, or near
the spinal cord; or in, on, or near a peripheral nerve (sensory or
motor) for the purpose of selective stimulation to achieve a
therapeutic purpose.
[0047] The distal end of the lead 12 may carry one or more
electrodes 14 to apply electrical pulses to the targeted tissue
region. The electrical pulses may be supplied by a pulse generator
16 coupled to the lead 12.
[0048] In the illustrated embodiment, the pulse generator 16 may be
implanted in a suitable location remote from the lead 12, e.g., in
the shoulder region. It should be appreciated, however, that the
pulse generator 16 may be placed in other regions of the body,
i.e., implanted in any suitable location, or externally.
[0049] When implanted, the case of the pulse generator 16 may serve
as a reference or return electrode. Alternatively, the lead 12 may
include a reference or return electrode (comprising a bi-polar
arrangement), or a separate reference or return electrode may be
implanted or attached elsewhere on the body (comprising a
mono-polar arrangement).
[0050] The pulse generator 16 may include an on-board, programmable
microprocessor 18, which carries embedded code. The code may
express pre-programmed rules or algorithms under which a desired
electrical stimulation waveform pattern or train is generated and
distributed to the electrode(s) 14 on the lead 12. According to
these programmed rules, the pulse generator 16 may direct the
prescribed stimulation waveform patterns or trains through the lead
12 to the electrode(s) 14, which serve to selectively stimulate the
targeted tissue region. The code may be preprogrammed by a
clinician to achieve the particular physiologic response
desired.
[0051] In the illustrated embodiment, an on-board battery 20 may
supply power to the microprocessor 18. Currently, batteries 20 must
be replaced every 1 to 9 years, depending on the stimulation
parameters needed to treat a disorder. When the battery life ends,
the replacement of batteries requires another invasive surgical
procedure to gain access to the implanted pulse generator. As will
be described, the system 10 makes possible, among its several
benefits, an increase in battery life.
[0052] The stimulation waveform pattern or train generated by the
pulse generator differs from convention pulse patterns or trains in
that the waveform comprises repeating non-regular (i.e., not
constant) pulse patterns or trains, in which the interval between
electrical pulses (the inter-pulse intervals or IPI) changes or
varies over time. Examples of these repeating non-regular pulse
patterns or trains are shown in FIGS. 3 to 5. Compared to
conventional pulse trains having regular (i.e., constant)
inter-pulse intervals (as shown in FIG. 2), the non-regular (i.e.,
not constant) pulse patterns or trains provide a lower average
frequency for a given pulse pattern or train, where the average
frequency for a given pulse train (expressed in hertz or Hz) is
defined as the sum of the inter-pulse intervals for the pulse train
in seconds (.SIGMA..sub.IPI) divided by the number of pulses (n) in
the given pulse train, or (.SIGMA..sub.IPI)/n. A lower average
frequency makes possible a reduction in the intensity of side
effects, as well as an increase in the dynamic range between the
onset of the desired clinical effect(s) and side effects, thereby
increasing the clinical efficacy and reducing sensitivity to the
position of the electrode(s). A lower average frequency brought
about by a non-regular pulse pattern or train also leads to a
decrease in power consumption, thereby prolonging battery life and
reducing battery size.
[0053] The repeating non-regular (i.e., not constant) pulse
patterns or trains can take a variety of different forms. For
example, as will be described in greater detail later, the
inter-pulse intervals can be linearly cyclically ramped over time
in non-regular temporal patterns (growing larger and/or smaller or
a combination of each over time); or be periodically embedded in
non-regular temporal patterns comprising clusters or groups of
multiple pulses (called n-lets), wherein n is two or more. For
example, when n=2, the n-let can be called a doublet; when n=3, the
n-let can be called a triplet; when n=4, the n-let can be called a
quadlet; and so on. The repeating non-regular pulse patterns or
trains can comprise combinations of single pulses (called singlets)
spaced apart by varying non-regular inter-pulse intervals and
n-lets interspersed among the singlets, the n-lets themselves being
spaced apart by varying non-regular inter-pulse intervals both
between adjacent n-lets and between the n pulses embedded in the
n-let. If desired, the non-regularity of the pulse pattern or train
can be accompanied by concomitant changes in waveform and/or
amplitude, and/or duration in each pulse pattern or train or in
successive pulse patterns or trains.
[0054] Each pulse comprising a singlet or imbedded in an n-let in a
given train comprises a waveform that can be monophasic, biphasic,
or multiphasic. Each waveform possesses a given amplitude
(expressed, e.g., in amperes) that can, by way of example, range
from 10 pa (E.sup.-6) to 10 ma (E.sup.-3). The amplitude of a given
phase in a waveform can be the same or differ among the phases.
Each waveform also possesses a duration (expressed, e.g., in
seconds) that can, by way of example, range from 10 .mu.s
(E.sup.-6) to 2 ms (E.sup.-3). The duration of the phases in a
given waveform can likewise be the same or different. It is
emphasized that all numerical values expressed herein are given by
way of example only. They can be varied, increased or decreased,
according to the clinical objectives.
[0055] When applied in deep brain stimulation, it is believed that
repeating stimulation patterns or trains applied with non-regular
inter-pulse intervals can regularize the output of disordered
neuronal firing, to thereby prevent the generation and propagation
of bursting activity with a lower average stimulation frequency
than required with conventional constant frequency trains, i.e.,
with a lower average frequency than about 100 Hz. FIG. 3 shows a
representative example of a repeating non-regular pulse pattern or
train in which the inter-pulse intervals are linearly cyclically
ramped over time. As shown in FIG. 3, the pulse pattern or train
includes singlet pulses (singlets) spaced apart by progressively
increasing inter-pulse intervals providing a decrease in frequency
over time, e.g., having an initial instantaneous frequency of 140
Hz, decreasing with doubling inter-pulse intervals, to a final
instantaneous frequency of 40 Hz. The inter-pulse intervals can
vary within a specified range selected based upon clinical
objections, e.g., not to exceed 25 ms, or not to exceed 100 ms, or
not to exceed 200 ms, to take into account burst responses and
subsequent disruption of thalamic fidelity. The non-regular pulse
trains repeat themselves for a clinically appropriate period of
time. As shown in FIG. 3, the first pulse train comprises
progressively increasing inter-pulse intervals from smallest to
largest, followed immediately by another essentially identical
second pulse train comprising progressively increasing inter-pulse
intervals from smallest to largest, followed immediately by an
essentially identical third pulse train, and so on. Therefore,
between successive pulse trains, there is an instantaneous change
from the largest inter-pulse interval (at the end of one train) to
the smallest inter-pulse interval (at the beginning of the next
successive train). The train shown in FIG. 3 has an average
frequency of 85 Hz and is highly non-regular, with a coefficient of
variation (CV) of about 0.5. As is demonstrated in the following
Example (Batch 3), the increased efficiency of the pulse train
shown in FIG. 3 (due to the lower average frequency) also can
provide greater efficacy, as compared to a constant 100 Hz pulse
pattern.
[0056] The train shown in FIG. 3 exploits the dynamics of burst
generation in thalamic neurons. The early high frequency phase of
the train masks intrinsic activity in subthalamic nucleus (STN)
neurons, and the inter-pulse interval increases reduce the average
frequency. A family of trains can be provided by varying the
initial frequency, final frequency, and rate of change within the
train, with the objective to prevent thalamic bursting with a lower
average stimulation frequency than required with constant frequency
trains.
[0057] FIGS. 4 and 5 show other representative examples of
repeating non-regular pulse patterns or trains. The pulse trains in
FIGS. 4 and 5 comprise within, a single pulse train, a combination
of single pulses (singlets) and embedded multiple pulse groups
(n-lets), with non-regular inter-pulse intervals between singlets
and n-lets, as well as non-regular inter-pulse intervals within the
n-lets themselves. The non-regular pulse trains repeat themselves
for a clinically appropriate period of time.
[0058] The non-regular pulse train can be characterized as
comprising one or more singlets spaced apart by a minimum
inter-pulse singlet interval and one or more n-lets comprising, for
each n-let, two or more pulses spaced apart by an inter-pulse
interval (called the "n-let inter-pulse interval") that is less
than the minimum singlet inter-pulse interval. The n-let
inter-pulse interval can itself vary within the train, as can the
interval between successive n-lets or a successive n-lets and
singlets. The non-regular pulse trains comprising singlets and
n-lets repeat themselves for a clinically appropriate period of
time.
[0059] In FIG. 4, each pulse train comprises four singlets in
succession (with non-regular inter-pulse intervals there between);
followed by four doublets in succession (with non-regular
inter-doublet pulse intervals there between and non-regular
inter-pulse intervals within each n-let); followed by a singlet,
three doublets, and a singlet (with non-regular inter-pulse
intervals there between and non-regular inter-pulse intervals
within each n-let). The temporal pattern of this pulse train
repeats itself in succession for a clinically appropriate period of
time. The non-regular temporal pulse pattern shown in FIG. 4 has an
average frequency of 67.82 Hz without loss of efficacy, as is
demonstrated in the following Example, Batch 17.
[0060] In FIG. 5, each pulse train comprises four singlets in
succession (with non-regular inter-pulse intervals there between);
followed by three doublets in succession (with non-regular
inter-doublet pulse intervals there between and non-regular
inter-pulse intervals within each n-let). The temporal pattern of
this pulse train repeats itself in succession for a clinically
appropriate period of time. The non-regular temporal pulse pattern
shown in FIG. 5 has an average frequency of 87.62 Hz without loss
of efficacy, as is demonstrated in the following Example, Batch
18.
[0061] The following Example illustrates a representative
methodology for developing and identifying candidate non-regular
stimulation trains as shown in FIGS. 3 to 5 that achieve comparable
or better efficacy at a lower average frequency (i.e., more
efficiency) than constant inter-pulse interval trains.
EXAMPLE
[0062] Computational models of thalamic DBS (McIntyre et al. 2004,
Birdno, 2009) and subthalamic DBS (Rubin and Terman, 2004) can be
used with genetic-algorithm-based optimization (Davis, 1991) (GA)
to design non-regular stimulation patterns or trains that produce
desired relief of symptoms with a lower average stimulation
frequency than regular, high-rate stimulation. McIntyre et al.
2004, Birdno, 2009; Rubin and Terman, 2004; and Davis, 1991 are
incorporated herein by reference.
[0063] In the GA implementation, the stimulus train (pattern) is
the chromosome of the organism, and each gene in the chromosome is
the IPI between two successive pulses in the train. The
implementation can start, e.g., with trains of 21 pulses (20 genes)
yielding a train length of .about.400 ms (at average frequency of
50 Hz), and the 6 s trains required for stimulation are built by
serial concatenation of 15 identical pulse trains. The process can
start with an initial population of, e.g., 50 organisms,
constituted of random WI's drawn from a uniform distribution. At
each step (generation) of the GA, the fitness of each pulse train
is evaluated using either the TC or basal ganglia network model
(identified above) and calculating a cost function, C. From each
generation, the 10 best stimulus trains (lowest C) are selected, to
be carried forward to the next generation. They will also be
combined (mated) and random variations (mutations) introduced into
the 40 offspring, yielding 50 trains in each generation. This
process assures that the best stimulation trains (traits) are
carried through to the next generation, while avoiding local minima
(i.e., mating and mutations preserve genetic diversity). See
Grefenstette 1986. The GA continues through successive generations
until the median and minimum values of the cost function reach a
plateau, and this will yield candidate trains.
[0064] The objective is to find patterns of non-constant
inter-pulse interval deep brain stimulation trains that provide
advantageous results, as defined by low frequency and low error
rate. An error function is desirably created that assigns the
output of each temporal pattern of stimulation a specific error
fraction (E) based on how the voltage output of the thalamic cells
correspond to the timing of the input stimulus. Using this error
fraction, a cost function (C) is desirably created to minimize both
frequency and error fraction, according to representative equation
C=W*E+K*f, where C is the cost, E is the error fraction, f is the
average frequency of the temporal pattern of stimulation, W is an
appropriate weighting factor for the error function, and K is an
appropriate weighting factor for the frequency. The weighting
factors W and K allow quantitative differentiation between efficacy
(E) and efficiency (f) to generate patterns of non-constant
inter-pulse interval deep brain stimulation trains that provide
advantageous results with lower average frequencies, compared to
conventional constant frequency pulse trains.
[0065] With this cost function, the voltage output of several
candidate temporal patterns of stimulation can be evaluated and the
cost calculated. Temporal patterns of stimulation with a low cost
can then be used to create new temporal patterns of similar
features in an attempt to achieve even lower costs. In this way,
new temporal patterns of stimulation can be "bred" for a set number
of generations and the best temporal patterns of stimulation of
each batch recorded.
[0066] Several batches of the genetic algorithm yields useful
results in that they achieve lower costs than the corresponding
constant frequency DBS waveforms. Some batches can be run in an
attempt to find especially low frequency temporal patterns of
stimulation, by changing the cost function to weight frequency more
heavily, or vice versa (i.e., by changing W and/or K). These
batches can also yield lower cost results than the
constant-frequency waveforms.
[0067] By way of example, a total of 14 batches of the genetic
algorithm were run and evaluated with various cost functions and
modified initial parameters.
[0068] Before the trials were run, a baseline was established by
running constant-frequency patterns of stimulation through the
model and analyzing the associated error fractions (Example FIG.
1). As can be seen from Example FIG. 1 (FIG. 6), the healthy
condition produced a low error fraction of 0.1 while the
Parkinsonian condition without DBS yielded a higher error fraction
of 0.5. From these results, constant high-frequency patterns of
stimulation ranging from 100-200 Hz gave near perfect results.
Novel non-constant temporal patterns of stimulation would then be
considered advantageous if they showed error fractions very close
to 0.1 with average frequencies less than 100-200 Hz.
[0069] The first set of batches was run by minimizing only the
error fraction (E). Thus, the associated cost function was simply
C=E. The results are summarized according to average frequency and
error fraction (Example Table 1). The associated inter-pulse
intervals (IPI's) can be seen in Example FIG. 2 (FIG. 7). Batch 3
outputted an error fraction 0.054. Another feature is that the
IPI's in Batch 3 gradually increased until about 40 msec, and then
repeated itself. This provides support that ramp trains are
advantageous. The trace shown in FIG. 3 generally incorporates the
temporal features of Batch 3.
[0070] The remaining batches yielded error fractions higher than
0.1 and were no better than the 150 Hz constant-frequency case.
TABLE-US-00001 EXAMPLE TABLE 1 Error Fraction Only, C = E # Average
Frequency Error Fraction IPI Length 3 127.5 0.054 5 4 95.62 0.162
39 5 113.6 0.139 13 6 94.64 0.132 26 7 101.6 0.142 31
[0071] Because many batches were yielding error fractions above 0.1
(healthy condition), and only a small window of error fraction less
than the 150 Hz DBS case would be useful, a new cost function was
constructed to minimize an alternate feature of the temporal
patterns of stimulation; namely, frequency. This new cost function
weighted the error fraction and frequency, yielding the equation
C=1000*E+F, where C is cost, E is error fraction, and F is the
average frequency of the waveform in Hz, W=1000, and K=1.
[0072] In order to establish a new baseline cost, the
constant-frequency patterns of stimulation were evaluated again
according to the new cost function (Example FIG. 3-FIG. 8). As can
be seen from the graph, the healthy condition reported a cost of
90.65 and the Parkinson case with no DBS yielded 505.50. The best
constant-frequency pattern of stimulation with the new cost
function was the 100 Hz case with a cost of 231.11. This new cost
function allowed for a wider range of solutions, because a temporal
pattern of stimulation would be considered useful if it had a cost
less than 231.11 but presumably higher than 90.65.
[0073] The results of the new cost function can be seen in Example
Table 2 and the IPI's visualized in Example FIG. 4 (FIG. 9). The
best results were seen in batches 15 and 18, which had the lowest
costs. Batch 18 also exhibits a ramp-like pattern of increasing
interpulse intervals. It shows a steadily falling IPI, followed by
a sudden rise, and then a quick fall, rise, and fall--almost as if
it consists of 3 smaller ramps. The trace shown in FIG. 5 generally
incorporates the temporal features of Batch 18. Batch 15 also
performed very well, but its qualitative features are more
difficult to discern.
TABLE-US-00002 EXAMPLE TABLE 2 Cost Function, C = 1000*E + F #
Average Frequency IPI Length Error Fraction Cost 9 94.74 34 0.124
218.8 13 132.9 12 0.087 219.4 15 98.00 17 0.098 196.0 18 81.28 10
0.116 197.3 19 84.70 20 0.116 201.2
[0074] The advantage of low frequency was emphasized with a new
cost function, which weighted frequency more heavily, C=1000*E+2*F.
Because the frequency of DBS does not affect the healthy condition
or the PD with no DBS, these baseline costs stayed the same at
90.65 and 505.50, respectively. The 100 Hz was again the best
constant-frequency temporal pattern of stimulation, with a cost of
331.11. The following temporal patterns of stimulation, then, were
considered useful if they had low frequencies and costs less than
331.11 and greater than 90.65.
[0075] The results of the revised cost function can be seen in
Example Table 3 and the IPI's visualized in Example FIG. 5 (FIG.
10). Of the resulting batches, batch 17 proved most interesting
because of its very low average frequency of 67.82 Hz. Even with
such a low frequency, it managed to prove better than the 100 Hz
condition with a reduction in cost of about 10. The waveform of
batch 17 is interesting in that it consists of a ramp pattern of
decreasing IPI in the first 100 msec, followed by a continual shift
between large IPI and small IPI. The qualitative feature of quickly
changing between large and small IPI's may prove advantageous. The
trace shown in FIG. 4 generally incorporates the temporal features
of Batch 17.
TABLE-US-00003 EXAMPLE TABLE 3 Revised Cost Function, Cost = 1000*E
+ 2*F # Average Frequency IPI Length Error Fraction Cost 16 84.92
47 0.239 323.8 17 67.82 20 0.253 321.1 20 79.25 10 0.236 315.4 21
77.15 20 0.269 346.6
[0076] The most interesting temporal patterns of stimulation in
this Example are from batches 15, 17, and 18. Batch 15 produced a
temporal pattern of stimulation with an average frequency of 98 Hz
with an error fraction as low as 0.098. Thus, it outperformed the
100 Hz constant-frequency case by managing to lower the error even
further at roughly the same frequency. Still, the qualitatively
useful features of batch 15 are difficult to discern. Batch 17 was
also appealing because of its very low frequency of 67.82. This low
frequency was gained at the cost of increased error at 0.253, but
it may nonetheless be useful if emphasis is placed on maintaining
low frequency DBS. The qualitative features of batch 17 indicated
at first a ramp followed by a continual switching between low and
high IPI's. Lastly, batch 18 stood somewhere in the middle with a
fairly low frequency of 87.62 and low error fraction of 0.116, only
marginally higher than the healthy condition of 0.1. The dominant
qualitative feature of batch 18's waveform is that it too shows a
ramp nature in that the IPI initially steadily falls, then quickly
rises, falls, and then rises. The rapid changing between high and
low IPI of batch 17 can be envisioned as a set of steep ramps.
[0077] A comparison of Batch 17 (FIG. 4) and Batch 18 (FIG. 5)
demonstrates how the balance between efficacy (as a function of the
model error fraction E) and efficiency (as a function of frequency
f) in non-regular temporal patterns of stimulation can be
purposefully tailored to meet clinical objectives. The systems and
methodologies discussed allow changing the cost function by
weighting efficacy or frequency more heavily (i.e., by changing W
and/or K), while still yielding temporal patterns of stimulation
with lower cost results than the constant-frequency waveforms.
Comparing Batch 17 with Batch 18, one sees that the error fraction
(E) (i.e., the efficacy of the temporal pattern) of Batch 17
(0.253) is greater than the error fraction (E) (i.e., the efficacy
of the temporal pattern) of Batch 18 (0.116). However, one can also
see that the efficiency (i.e., the average frequency) of Batch 17
(67.82 Hz) is lower than the efficiency (i.e., the average
frequency) of Batch 18 (81.28 Hz). Through different in terms of
efficacy and efficiency, both Batch 17 and Batch 18 have costs
better than constant-frequency temporal patterns.
[0078] FIG. 11 depicts a modeled raster of healthy firing of
neurons in the subthalamic nucleus and the global pallidus, both
external and internal segments thereof, through about one second of
time.
[0079] FIG. 12 depicts a modeled raster of the healthy subject
modeled in FIG. 11, with the addition of a forced Parkinsonian
state, however with deep brain stimulation not being applied to the
model. Like the raster of FIG. 11, this Figure depicts neurons in
the subthalamic nucleus and the global pallidus, both external and
internal segments thereof, through about one second of time.
[0080] FIG. 13 depicts a modeled raster of the modeled Parkinsonian
subject of FIG. 12, however further applying a regular 45 Hz
regular interval DBS signal to the subthalamic nucleus, as can be
seen. Additionally, the figure shows the firing of neurons in the
global pallidus, both external and internal segments thereof,
through about one second of time.
[0081] FIG. 14 depicts a modeled raster of the modeled Parkinsonian
subject of FIG. 12, however further applying a regular 100 Hz
regular interval DBS signal to the subthalamic nucleus, as can be
seen. Additionally, the figure shows the firing of neurons in the
global pallidus, both external and internal segments thereof,
through about one second of time.
[0082] FIG. 15 depicts a graphical representation of modeled,
conventionally expected error fractions where a regular interval
DBS signal is applied to the Parkinsonian model. The goal in
determining optimum stimulation patterns may be to provide a
stimulation pattern that has a lower average frequency with at
least as good, if not better (lower) error fractions than regular
interval DBS signals presented to the STN. By way of a non-limiting
example, with DBS off (average frequency=0 Hz), the model provides
that an expected error fraction may be about 0.34 to about 0.40.
With a regular interval stimulation pattern applied to the STN at
about 45 Hz, the expected error fraction is about 0.20 to about
0.25. According to the model, and therefore generally accepted in
the field, higher average frequency regular interval stimulation
yields a lower error fraction. Accordingly, if stimulation could be
provided at an average frequency of about 45 Hz with a modeled
error fraction less than that expected (i.e., less than about
0.20), benefits would be realized. Not only would relief from brain
disorders be improved, but power consumption by any device
delivering the new stimulation patterns, would be reduced as
compared to the same device delivering regular interval stimulation
patterns in an attempt to achieve similar performance results.
[0083] FIG. 16 provides an illustration of a model structure that
may be used to generate stimulation patterns according to the
present invention. Reference to this illustration may be helpful in
explaining what is referred to herein as an "error fraction." As
used herein, an "error fraction" is generally understood to mean
the number of errors occurring at the output of a model as compared
to the number of inputs provided to the model. An output error
occurs when a contrast arises between an expected value of the
model output to an output generated by the model provided with a
given stimulation pattern, such as to the STN, as shown.
[0084] FIG. 17 generally depicts a known genetic algorithm process
model, beginning with the generation of organisms (in this case
pulse trains or stimulation patterns), and continuing as described
above. One method for a mating process that may be employed in the
genetic algorithm according to the present invention is a single
crossover process by which certain, but preferably not all, genes
(stimulation pulses) are exchanged between parent stimulation
trains so as to yield two child stimulation trains are generated.
As depicted in FIG. 18, the stimulation patterns include a series
of 1's and 0's, which indicate whether or not, respectively, a
stimulation pulse is to be delivered during a given time step, such
as about 500 microseconds to about 100 milliseconds, and preferably
about one to 5 milliseconds. While initial or starting stimulation
patterns may be created by drawing interpulse intervals from some
distribution, such as a Gaussian distribution, the initial
stimulation patterns are preferably generated randomly, and
constraints may be added to control the number of stimulation
pulses (1's) in the initial stimulation patterns, thereby
controlling the average frequency range of the stimulation pattern.
Resulting generational stimulation patterns are then evaluated by
the model and compared to the performance of regular interval
stimulation patterns provided to the model at the same average
frequency as the generational stimulation pattern currently under
evaluation.
[0085] Another cost function that has proven useful in determining
beneficial non-regular temporal patterns of stimulation generated
by a genetic algorithm is as follows:
C=(E.sub.GA-E.sub.FMReg)/E.sub.FMReg*100% where E.sub.GA is the
error fraction of a selected generational stimulation pattern
generated by the genetic algorithm and currently under analysis by
the model and E.sub.FMReg is the error fraction of a DBS
stimulation pattern of uniform frequency at a frequency equal to
the average frequency of the GA train under analysis. This may be
referred to generally as a percent change cost function. At first,
one might expect that this cost function would not force a genetic
algorithm to search for non-regular patterns of DBS with a low
average frequency. However, this is not the case; the GA is
inclined to search for non-regular stimulation pattern of DBS with
a low average frequency because there is a greater opportunity to
find improved stimulation patterns (i.e., having a lower error
fraction) at lower frequencies. That is, as shown in FIG. 15, the
error fraction associated with 130 Hz conventional regular-interval
DBS is already so close to zero that it is highly unlikely that a
non-regular pattern with an average frequency of 130 Hz is going to
have a smaller error fraction. In other words, there just is not
much room to improve at 130 Hz. On the other hand, at 45 Hz, there
is ample room for improvement. It is much more likely that a
non-regular pattern of DBS with an average frequency of 45 Hz will
be found that has a better performance than conventional
regular-interval DBS provided at 45 Hz. Therefore, using a percent
decrease cost function implicitly incorporates the average
frequency of DBS while helping to minimize complications of
selecting weighting parameters, as with the other cost functions
discussed herein. As shown in FIG. 19, it is beneficial to run the
genetic algorithm through a plurality of generations so as to
decrease the cost. This figure shows the (decline in) cost as the
genetic algorithm progresses. That is, the algorithm is identifying
better and better stimulation patterns, from generation to
generation, and subsequently the cost is declining, or, in other
words, the performance is increasing.
[0086] Three stimulation trains generated according to the present
invention utilizing the percent change cost function to guide
survival and propagation are shown in FIGS. 20-22. In FIG. 20, a
preferred GA1 stimulation train is shown as being applied to the
STN of a Parkinsonion brain model. The preferred stimulation
pattern includes the repetition of a set 100 of three triplet
stimulation pulses 100A,100B,100C, where each triplet preferably
comprises a singlet 102 followed by a doublet 104. Through any
given set 100 of triplets, the interpulse interval between the
singlet and doublet of one triplet is preferably different than the
interpulse interval between the singlet and doublet of at least one
other triplet, and more preferably different than each interpulse
interval between the singlet and doublet of all other triplets in
the set 100. Furthermore, the interpulse interval within the
doublets of each triplet is preferably different than at least one
other doublet interpulse interval within that set 100, and more
preferably the interpulse interval within each doublet in a given
set 100 is different than the interpulse interval within each other
doublet in that set 100. While the interpulse intervals between the
singlets and doublets of a given set, and within the doublets of
the given set, may vary (as much as integer factors), the
interpulse interval between each triplet within a given set
preferably remains relatively constant, such as by varying less
than about 10% throughout the set.
[0087] In each of the described preferred stimulation patterns, a
given set to be repeated includes at least one doublet and at least
one singlet. As in the case of the GA1 train, the number of
singlets and doublets in a given set 100 was equal (three of each).
As in the case of the GA2 pattern set 200, as shown in FIG. 21, the
number of doublets far outweighed the number of singlets (80% of
n-lets were doublets as opposed to 20% as singlets). As in the case
of the GA3 pattern set 300, as shown in FIG. 22, the number of
doublets outweighed the number of singlets (60% of n-lets were
doublets as opposed to 40% as singlets). Accordingly, it may be
more preferable to include, in a given stimulation pattern to be
repeatedly delivered to the subthalamic nucleus or other portion of
the brain, one or more singlets and one or more doublets, where the
number of doublets is equal to or greater than the number of
singlets in the set.
[0088] As shown in FIG. 23, the stimulation patterns generated
according to the present invention have a much lower error fraction
value as compared to their regular interval stimulation
counterparts at frequencies less than 100 Hz. For example, at 45
Hz, regular interval DBS has an error fraction of about 20 in 100
(0.20) to about 25 in 100 (0.25). On the other hand, the
stimulation patterns generated according to the present invention
provide a modeled error fraction of about 5 in 100 to about 15 in
100, thereby demonstrating a 25% to 80% improvement in
efficiency.
[0089] An electrical stimulation pattern created according to the
present invention (GA1) was experimentally applied to two human
patients that had been diagnosed with Parkinson's Disease. The GA1
pattern was applied during intraoperative experiments that were
conducted by connecting to an exposed lead of a previously
implanted DBS electrode during an implantable pulse generator
replacement surgery. After connection, the GA1 pattern of
stimulation and a few control patterns were delivered. Motor
impairment was quantified while delivering the patterns of
stimulation using a known finger-tapping task. To measure the
effect of the DBS stimulation patterns according to the present
invention, a two-button computer mouse was utilized, and the
patient was instructed to, during data collection times, alternate
clicking a respective mouse button with their index finger and
their middle finger. The time duration of the respective button
clicks was then recorded by a computer and analyzed. The time
duration of one or both fingers may be analyzed, depending upon
statistical results. As can be seen in FIG. 24, the GA1 stimulation
pattern allowed each patient to demonstrate an increase in motor
function as compared to the regular interval DBS stimulation
pattern provided at the same average frequency, thus indicating an
increased benefit in performance with no sacrifice to average cost
(i.e., no increase in average power). Furthermore, for Patient 1,
the GA1 stimulation pattern caused the patient to perform
substantially similar to motor function demonstrated under
application of a regular interval DBS stimulation pattern of 185
Hz, thus indicating substantial similar performance with a great
cost (i.e. power) reduction (stimulation provided at an average of
45 Hz instead of 185 Hz). Finally, with respect to Patient 2, the
GA1 stimulation pattern caused the patient to perform better than
the motor function demonstrated under application of a regular
interval DBS stimulation pattern of 185 Hz, thus indicating
improved performance with a great cost (i.e. power) reduction
(stimulation provided at an average of 45 Hz instead of 185 Hz).
Generally speaking, then, the 45 Hz average frequency stimulation
pattern designed according to the present invention performed
similarly or better than conventional 185 Hz regular DBS
stimulation and better than frequency matched (45 Hz) regular
stimulation. Further clinical experiments have been conducted in
applying stimulation to the STN using stimulation patterns
generated according to the present invention, and such experiments
show promising results.
[0090] An electrical stimulation pattern created according to the
present invention (GA1) was also tested in one human subject with
Parkinson's disease where tremor was that subject's primary motor
symptom. The subject's tremor was quantified using as accelerometer
on the back of the subject's contralateral wrist.
[0091] Tremor was measured in the contralateral limb during
unilateral stimulation with a temporal pattern of stimulation
generated according to the present invention having an average
frequency of about 45 Hz, regular 45 Hz and 185 Hz stimulation, and
with stimulation off (controls) in a single intraoperative session
with a human subject. The stimulation pattern was presented to the
subject, and the subject was blinded to the experimental condition.
The trial began with one minute of stimulation off, with baseline
tremor measured for 20 seconds beginning about 30 seconds into
these intervals, and about 30 seconds after each condition was
initiated experimental tremor was measured for 20 seconds (Ex. FIG.
28A).
[0092] Tremor was measured using an accelerometer (Crossbow
CXL04LP3; 5V/4g sensitivity, San Jose, Calif., USA) taped to the
dorsum of the hand. The amplitude of tremor recorded by an
accelerometer generally correlates well with clinical tremor rating
scales. To obtain a single quantitative measure of tremor, the
power spectral density was calculated for each of the three
acceleration signals (AX, AY, and AZ, Ex. FIG. 28B) using the psd
function (power spectral density, Welch's averaged periodogram,
Hanning window, FFT length=5,000) in MATLAB (Mathworks Inc.,
Natick, Mass., USA). Next, we integrated each spectrum from 2-20 Hz
to get PX, PY, and PZ. Finally, we summed PX, PY, and PZ, and took
the log of the sum to get a single metric of tremor. The frequency
range of 2-20 Hz was chosen to include the primary and several
harmonics of the tremor and to exclude steady state acceleration
due to gravity.
[0093] As mentioned, the power spectral density for the
acceleration signal was integrated from 2-20 Hz in order to get a
single quantitative measure of the tremor amplitude. As can be seen
in FIG. 25, the stimulation pattern generated according the present
invention having an average frequency of about 45 Hz (GA1) reduced
the tremor amplitude more than the regular 45 Hz stimulation but
slightly less than regular 185 Hz stimulation. All three patterns
of stimulation reduced tremor amplitude compared to the stimulation
off condition.
[0094] FIG. 26 provides a spectral analysis of the average power
across 10 GPi neurons for a single stimulation pattern iteration
applying the indicated stimulation pattern to a forced Parkinsonian
state in a model. As can be seen, there is significant oscillatory
or synchronous activity generated around 15 Hz in the Parkinsonian
state when the DBS input is off. The 45 Hz regular interval
stimulation does not lessen such activity much, but the 185 Hz
regular interval stimulation does. Accordingly, there may be a
correlation between an attenuation of such oscillatory or
synchronous activity and the effectiveness of a given DBS
stimulation pattern. Indeed, it has been observed that such
attenuation is at least correlated to an improvement in movement,
especially in animals that have a previously demonstrated or
induced state of bradykinesia.
[0095] FIG. 27 provides an example tuning curve to analyze the
10-30 Hz global pallidus internal (GPi) neuronal power. That is,
the spectra for 10 GPi neurons in a model were averaged for each
stimulation state across 10 iterations. The plotted points indicate
the average power of the GPi spike times across all 10 GPi neurons,
averaged across the 10 stimulation iterations at each stimulation
state. As can be seen, the power demonstrated in the 10-30 Hz
frequencies by the GPi neurons (which are generally correlated to a
neurological condition) is greatly reduced by the application of
stimulation patterns that have been clinically shown to assist in
reducing negative effects of such conditions. Accordingly,
stimulation patterns according to the present invention may be, and
have found to be, directed towards or involve the reduction of the
average power of GPi oscillatory or synchronous activity. In fact,
as shown in FIG. 27, an attenuation of at least about 25% of
oscillatory activity caused by an actual animal neurological
condition or a neurological condition model, but more preferably an
attenuation of about 50% or greater, may be achieved using
stimulation patterns according to the present invention.
[0096] The power of the oscillatory or synchronous activity that
may be modeled, or measured from a patient, as correlated to a
neurological condition may be used in alternative cost functions
according to another embodiment of the present invention for
optimizing stimulation patterns. One cost function that may be
employed by an optimization algorithm or technique according to the
present invention is as follows:
C=(P.sub.GA-P.sub.FMReg)/P.sub.FMReg*100% where PGA is the average
power generated by a computer model, over a given frequency range,
of the firing of one or more GPi neurons when a selected
generational stimulation pattern, which was initially created or
generated by the genetic algorithm, is applied to the STN in the
model and PFMReg is the power generated by a computer model, over
the same given frequency range, of the firing the same GPi neurons
when a DBS stimulation pattern of uniform frequency at a frequency
equal to the average frequency of the GA train under analysis. The
given frequency range may be a single frequency (e.g. 15 Hz) or a
set of preferably contiguous frequencies (e.g. 10-30 Hz) or a set
of noncontiguous frequencies (e.g. 15, 20, and 30 Hz).
[0097] Another cost function, using oscillatory power, that may be
used to optimize stimulation patterns is as follows: C=W*P+K*f,
where C is the cost, P is the average power generated by a computer
model, over a given frequency range, of the firing of one or more
GPi neurons when a selected generational stimulation pattern, which
was initially created or generated by the genetic algorithm, is
applied to the STN in the model, f is the average frequency of the
generational pattern of stimulation, W is an appropriate weighting
factor for the average power, and K is an appropriate weighting
factor for the frequency. The given frequency range may be a single
frequency (e.g. 15 Hz) or a set of preferably contiguous
frequencies (e.g. 10-30 Hz) or a set of noncontiguous frequencies
(e.g. 15, 20, and 30 Hz). The weighting factors W and K allow
quantitative differentiation between efficacy (as a function of P)
and efficiency (as a function of f) to generate patterns of
non-constant inter-pulse interval deep brain stimulation trains
that provide advantageous results with lower average frequencies,
compared to conventional constant frequency pulse trains.
[0098] The non-regular temporal patterns of stimulation generated
and disclosed above therefore make possible achieving at least the
same or equivalent (and expectedly better) clinical efficacy at a
lower average frequency compared to conventional constant-frequency
temporal patterns. The lower average frequencies of the non-regular
temporal stimulation patterns make possible increases in efficiency
and expand the therapeutic window of amplitudes that can be applied
to achieve the desired result before side effects are
encountered.
[0099] DBS is a well-established therapy for treatment of movement
disorders, but the lack of understanding of mechanisms of action
has limited full development and optimization of this treatment.
Previous studies have focused on DBS-induced increases or decreases
in neuronal firing rates in the basal ganglia and thalamus.
However, recent data suggest that changes in neuronal firing
patterns may be at least as important as changes in firing
rates.
[0100] The above described systems and methodologies make it
possible to determine the effects of the temporal pattern of DBS on
simulated and measured neuronal activity, as well as motor symptoms
in both animals and humans. The methodologies make possible the
qualitative and quantitative determination of the temporal features
of low frequency stimulation trains that preserve efficacy.
[0101] The systems and methodologies described herein provide
robust insight into the effects of the temporal patterns of DBS,
and thereby illuminate the mechanisms of action. Exploiting this
understanding, new temporal patterns of stimulation may be
developed, using model-based optimization, and tested, with the
objective and expectation to increase DBS efficacy and increase DBS
efficiency by reducing DBS side effects.
[0102] The invention provides non-regular stimulation patterns or
trains that can create a range of motor effects from exacerbation
of symptoms to relief of symptoms. The non-regular stimulation
patterns or trains described herein and their testing according to
the methodology described herein will facilitate the selection of
optimal surgical targets as well as treatments for new disorders.
The non-regular stimulation patterns or trains described herein
make possible improved outcomes of DBS by potentially reducing side
effects and prolonging battery life. The extended battery life will
result from a lower average frequency of stimulation (45 Hz vs. 100
or 185 Hz), thereby delivering less electrical current over time.
Surgeries to replace depleted pulse generators will be needed less
frequently and the costs that a DBS patient can expect with the DBS
system will be diminished.
[0103] The foregoing is considered as illustrative only of the
principles of the invention. Furthermore, since numerous
modifications and changes may readily occur to those skilled in the
art, it is not desired to limit the invention to the exact
construction and operation shown and described. While the preferred
embodiment has been described, the details may be changed without
departing from the invention. For instance, although the disclosed
embodiments of an algorithm used to generate stimulation patterns
is an evolutionary algorithm, namely a genetic algorithm, the scope
of the methods for this technology is not limited to genetic
algorithms. Indeed, the scope of the present invention includes
other contemplated model-based optimization techniques including,
but not limited to, other evolutionary algorithms, swarm
intelligence algorithms, and other optimization techniques or
metaheuristic. The scope of the present invention is not limited to
any particular model of a neurological disorder, such as PD.
Present or future models of neurological disorders that are treated
with DBS, or other electrical stimulation, are candidates for use
with the methods described herein. Furthermore, while certain
electrical stimulation patterns have been clinically applied in an
effort to quantify their efficacy and efficiency, it will be
appreciated that the scope of the present invention is not
necessarily limited to any particular stimulation pattern as
disclosed, but rather the scope of the present invention
encompasses all patterns generated according hereto. The claims as
follows are intended to include all modifications and alterations
insofar as they come within the scope of the claims or the
equivalent thereof
* * * * *